VIBRO-ACOUSTICAL SOUND REPRODUCTION in CARS Sebastian Merchel and M
Total Page：16
File Type：pdf, Size：1020Kb
VIBRO-ACOUSTICAL SOUND REPRODUCTION IN CARS Sebastian Merchel and M. Ercan Altinsoy TU Dresden, Chair of Communication Acoustics, Dresden, Germany email: [email protected], Dirk Kaule BMW Group Research and Technology, Munich, Germany Christian Volkmar IAV GmbH, Munich, Germany Usually, headphones are used for seat-related entertainment systems in passenger vehicles. The goal of this project is the individual multi-seat audio reproduction using only loudspeakers. The main challenge is the crosstalk between different positions at low frequencies. One approach to solve this problem is the substitution of airborne sound by seat vibrations. Previous studies have shown, that such vibrations can improve the perceived quality of the music experience [1, 2, 3, 4]. Additionally, vibrations do influence the perceived intensity of a sound at low frequencies - the so- called auditory-tactile loudness illusion [5, 6]. This article discusses if the individual acoustic bass level can be reduced by excitation of seat vibrations. A car seat equipped with multiple vibration actuators was developed for the experiments. Different perceptually optimized algorithms were applied to generate vibrations from audio signals, taking into account the transfer characteristics of the exciters. Several music sequences were selected as test stimuli. The influence of different vibration levels was also investigated. Introduction It has been shown in a previous study that vibrations (mechanical stimuli, which excite the surface of the body and stimulate the tactile sense) are able to influence the perceived loudness of a sound at low frequencies [6]. Sinusoidal tones were perceived to be louder when vibrations were reproduced simultaneously via a seat. Because of this cross-modal loudness illusion, it might be possible that the ideal low-frequency audio equalization might change under the influence of additional vibrations in a music reproduction scenario. Therefore, this study investigated if the bass level can be reduced under the influence of seat vibrations. Several music sequences were selected as test sequences. The influence of different vibration levels was also investigated. ICSV22, Florence, Italy, 12-16 July 2015 1 The 22nd International Congress of Sound and Vibration Setup To investigate the coupled perception of vibrations and sound, a reproduction system was devel- oped. Vibrations were generated via a car seat equipped with multiple vibration actuators. A first shaker was attached to the bottom of the seat and a second shaker behind the backrest. Figure 1 shows the seat and the position of the actuators. The transfer characteristic of the vibrating chair (acceleration at the surface of the seat versus voltage input) was measured for multiple subjects. A vi- bration pad (Brüel & Kjær Type 4515B) was used. The averaged transfer function (15 measurements: 3 subjects and 5 repetitions) at the seating is plotted in Figure 2. The characteristic at the backrest is comparable. A distinct resonance behavior can be seen. Inverse filters were designed in MATLAB to equalize the amplitude response. For real-time implementation in Pure Data, the resulting FIR filters (40th order) were divided into 20 biquad sections. Vibration generation was also implemented in Pure Data by simply low-pass filtering the mono sum of the audio signal. A 200 Hz Butterworth filter of 10th order was used. More advanced perceptually optimized algorithms, which will not be described here, were developed to generate vibrations from audio signals. For a discussion of the influence of various algorithms on the perceived concert quality in a music reproduction scenario please refer to [4]. Figure 1: Car seat with the position of electro-dynamic shakers. For audio reproduction a conventional stereo setup was placed in a sound studio. Two Genelec 8050A loudspeakers and a Genelec 7060B subwoofer were used. The loudspeakers were placed in ear height with a distance of 2.5 m. The system was equalized to a flat frequency response at the listener position in the center of the room. 2 ICSV22, Florence, Italy, 12-16 July 2015 The 22nd International Congress of Sound and Vibration -10 without compensation -15 with compensation -20 -25 / V] / V] 2 -30 -35 L / dB [m/s -40 -45 -50 40 50 6080 100 120 160 200 f / Hz Figure 2: Averaged transfer function (FFT 65536, 1/24th octave intensity averaging) measured at the seat surface of the vibration chair, with and without compensation. Subjects A total of 19 subjects participated in the experiment (13 male and 6 female). They were between 23 and 28 years old (mean 25 years). All participants stated that they had no known hearing or spine damage. Most of the test subjects had never participated in a listening experiment before. In general, the listeners can be regarded as naive, with no background in audio engineering or sound evaluation. Stimuli and Experimental Design Four music sequences were selected that include low-frequency content: – BRADLEY, Charles - You Put The Flame On It, from: Victim Of Love, Daptone / Dunham Records, 2013 – BEATSTEAKS - Hand In Hand, from: Smack Smash, Epitaph / Warner, 2004 – MODERAT - Bad Kingdom, from: II, Monkeytown Records, 2013 – Wir sind HELDEN - Gekommen um zu bleiben, from: Von hier an blind, EMI Music, 2005 The sequences from different genres varied in speed and type of the bass content. The sequence MODERAT contained synthetic electronic sounds. The other sequences were recorded with electric and acoustic bass instruments. Also the temporal structure of the bass line varied between stimuli. There were long bass notes (BEATSTEAKS) and fast bass lines (HELDEN). The loudness of the four sequences was equalized in a pilot experiment using two subjects. The resulting equivalent continuous sound pressure level was measured; for instance, a value of approximately 73 dB(A) was obtained for the sequence BRADLEY. The first seconds were removed from two sequences, because not much bass was contained in the audio signal (BEATSTEAKS - ca. 10 s, HELDEN - ca. 9 s). To ensure that the participants had sufficient time for their evaluations, the stimulus length was not limited. A diagram of a sample trial is shown in Figure 3. The reference interval without vibration and the test test interval with vibration were played back in turns. The participants were allowed to switch ICSV22, Florence, Italy, 12-16 July 2015 3 The 22nd International Congress of Sound and Vibration individually between both intervals at any time and as often as wanted. A computer screen provided feedback about the currently activated interval for better orientation. The reference interval contained the original unmodified music signal with fixed bass level (flat frequency response). In contrast, the bass level in the test interval was adjustable. The listeners were asked to adjust the bass level during the test interval, until the bass intensity in the reference and the test interval felt the same. The vibration in the test interval remained the same throughout each trial. Adjacent intervals were separated by 0.75 s breaks. The music sequence was continued at the same position at which it ended in the last interval. Reference TestReference Test ... Bass level variable 0.75s variable ... Vibration Figure 3: Experimental procedure: The test participant was allowed to switch between the reference and test interval until she/he was satisfied with her/his adjustment of the bass level. Throughout each trial, the vibration amplitude remained constant at a 0 dB, 3 dB, 6 dB, 9 dB or 12 dB sensation level. A shelving filter at 200 Hz was implemented in Pure Data for adjusting the bass level in real time. An infinite rotary knob with no visible marks was applied (PowerMate, Griffin Technology). Intensity increments were possible in 0.5 dB steps over a ±6 dB range. The initial bass level in each trial was randomly varied over the complete dynamic range. The listeners were asked to adjust the bass level for five levels of vibration. Additionally, a condi- tion without vibration was investigated. The acceleration levels referred to the individual perception threshold to ensure that the perceived vibration magnitude was comparable between subjects. The threshold was measured for each participant in a short test. Supra-threshold white noise was used as vibration signal. Masking white noise was reproduced throughout the experiment at 74 dB via head- phones to reduce the influence of possible sound radiation of the seat. The subject was asked to use the rotary knob described above, to reduce the vibration intensity until the vibrations were just per- ceivable. Subsequently, the threshold was tested again. The experimenter raised the intensity slowly until the subject raised her/his hand to indicate that she/he felt the vibration. An averaged perception threshold was calculated from the two measurements for each participant. The five acceleration levels were then adjusted relative to the individual thresholds and were fur- ther labeled as relative vibration levels: 0 dB (just perceivable vibration) , 3 dB, 6 dB (medium vibra- tion), 9 dB, 12 dB (strong vibration). A relative vibration level of 0 dB means that the averaged RMS acceleration level of a music sequence was equal to the acceleration level at the perception thresh- old. However, the peak acceleration level is slightly above this threshold - depending on the music sequence. To test the reliability of each subject, there were two repetitions of each sequence. This process resulted in a total of 48 trials (four sequences × six vibration levels × two repetitions). The listeners were allowed to stop a trial as soon as they were satisfied with their adjustment. Before the experi- mental session, the listeners were familiarized with the procedure and all of the music sequences. The overall test time varied between subjects and took up to 1:30 h including a break. 4 ICSV22, Florence, Italy, 12-16 July 2015 The 22nd International Congress of Sound and Vibration Results and Discussion Figure 4 presents the mean adjusted bass levels, for which the perceived bass intensity was judged as equal in both intervals.